Automated Registration Of LiDAR Image To Remote Sensing Image Based On Harris Corner

LiDAR data is the three-dimensional data of the earth's surface with high accuracy, we can obtain DSM(Digital Surface Model) and DTM(Digital Terrain Model) by automated post-processing techniques, and reconstruct three-dimensio- nal model of target for urban planning, environment monitoring, including pre-disaster forecasting, disaster and war emergency situations to provide possible rapid responses. However, horizontal resolution of LiDAR image is relatively low. This thesis studied on image registration between LiDAR images with low resolution and remote sensing images with high resolution to prepare for the surface three-dimensional model reconstruction with high accuracy.In this thesis, the main studies are as follows:(1) We summed up image registration research results and progress home and abroad in recent years, it mainly focused on feature points matching algorithms, and compared with the common feature point extraction operators to figure out the feature extraction operator with better accuracy and stability.(2) We combined the commonalities of LiDAR images and remote sensing images with their respective characteristics to fulfill two kinds of image registration algorithms, which are suitable for image registration between LiDAR images and remote sensing images. 1) Image registration method based on corner and mutual information. Firstly, we extracted corners from LiDAR image and remote sensing image, then we stood on the principle of mutual information maximization between corners to find out rotation among images and the translation. 2) Image registration method based on corner integrated with intensity mapping. Firstly, we extracted the Harris corner points in the remote sensing image, and then we searched for the optimal matched positions in LiDAR image using sum of squared differences combined with gray scale remapping when it got minimum. Finally, we implemented registration by using the right feature matched point pairs between the two images.(3) Experiments have verified the effectiveness of the image registration methods based on corner mutual information and based on corner integrated with intensity mapping toward LiDAR image and remote sensing image. Experiments have verified the robustness of the two image registration algorithms by adding noise to LiDAR image and remote sensing image.